2002-10-22 18:38:41

by Shaya Potter

Subject: [PATCH & RFC] first attempt at making chroot nestable

This is a first attempt at a change that lets kernel's chroot's and wont
let you use an fd to break out of a chroot, it will only let you get as
far as the root as of when you got the fd (or possibly even less).

I've discussed the algorithm I've implemented below a couple times
already on the list, but I doubt my coding style is up to snuff, so
criticize away on that. I also haven't had a chance to test it yet
(need to get into school to use the test machines), so all I know is
that it compiles w/o any warnings.

please comment away (as well as tell me if I did right my inline'ing the
patch, or if I should attach it, if it matters the mail client I
normally use is evolution)

thanks,

shaya

diff -Naur linux-2.4.19/fs/open.c linux-2.4.19.new/fs/open.c
--- linux-2.4.19/fs/open.c 2002-10-22 14:28:48.000000000 -0400
+++ linux-2.4.19.new/fs/open.c 2002-10-22 14:29:37.000000000 -0400
@@ -475,10 +475,12 @@
set_fs_root(current->fs, nd.mnt, nd.dentry);
set_fs_altroot();

+ write_lock(&current->fs->lock);
cp->mnt = mntget(nd.mnt);
cp->root = dget(nd.dentry);
cp->next = current->cp;
current->cp = cp;
+ write_unlock(&current>fs->lock);

error = 0;
dput_and_out:
diff -Naur linux-2.4.19/include/linux/sched.h linux-2.4.19.new/include/linux/sched.h
--- linux-2.4.19/include/linux/sched.h 2002-08-02 20:39:45.000000000 -0400
+++ linux-2.4.19.new/include/linux/sched.h 2002-10-22 13:54:12.000000000 -0400
@@ -418,6 +418,7 @@

/* journalling filesystem info */
void *journal_info;
+ struct chroot_point *cp;
};

/*
@@ -511,6 +512,7 @@
blocked: {{0}}, \
alloc_lock: SPIN_LOCK_UNLOCKED, \
journal_info: NULL, \
+ cp: NULL, \
}


diff -Naur linux-2.4.19/kernel/exit.c linux-2.4.19.new/kernel/exit.c
--- linux-2.4.19/kernel/exit.c 2002-08-02 20:39:46.000000000 -0400
+++ linux-2.4.19.new/kernel/exit.c 2002-10-22 13:54:09.000000000 -0400
@@ -286,6 +286,31 @@
__exit_fs(tsk);
}

+static inline void __exit_chroot(struct task_struct *tsk)
+{
+ struct chroot_point *cur, *next;
+
+ cur = tsk->cp;
+
+ task_lock(tsk);
+ tsk->cp = NULL;
+ task_unlock(tsk);
+
+ while (cur)
+ {
+ kfree(cur->mnt);
+ kfree(cur->root);
+ next = cur->next;
+ kfree(cur);
+ cur = next;
+ }
+}
+
+void exit_chroot(struct task_struct *tsk)
+{
+ __exit_chroot(tsk);
+}
+
/*
* We can use these to temporarily drop into
* "lazy TLB" mode and back.
@@ -459,6 +484,7 @@
sem_exit();
__exit_files(tsk);
__exit_fs(tsk);
+ __exit_chroot(tsk);
exit_namespace(tsk);
exit_sighand(tsk);
exit_thread();
diff -Naur linux-2.4.19/kernel/fork.c linux-2.4.19.new/kernel/fork.c
--- linux-2.4.19/kernel/fork.c 2002-08-02 20:39:46.000000000 -0400
+++ linux-2.4.19.new/kernel/fork.c 2002-10-22 14:15:38.000000000 -0400
@@ -580,6 +580,7 @@
int retval;
struct task_struct *p;
struct completion vfork;
+ struct chroot_point *cp_p, *cp_c;

if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
return -EINVAL;
@@ -726,6 +727,28 @@
p->tgid = retval;
INIT_LIST_HEAD(&p->thread_group);

+ /* copy chroot_point list from parent to child */
+ cp_p = current->cp;
+ cp_c = NULL;
+
+ if (cp_p)
+ cp_c = (struct chroot_point *) kmalloc(sizeof(struct chroot_point), GFP_KERNEL);
+
+ p->cp = cp_c;
+
+ while (cp_p)
+ {
+ cp_c->root = dget(cp_p->root);
+ cp_c->mnt = mntget(cp_p->mnt);
+ cp_c->next = NULL;
+
+ if (cp_p->next) {
+ cp_c->next = (struct chroot_point *) kmalloc(sizeof(struct chroot_point), GFP_KERNEL);
+ cp_c = cp_c->next;
+ }
+ cp_p = cp_p->next;
+ }
+
/* Need tasklist lock for parent etc handling! */
write_lock_irq(&tasklist_lock);